South Korea's Nuclear Comeback Is Driven by a 27GW Power Surge

The Number That Rewrote South Korea’s Energy Future

In April, South Korea projected its 2040 peak electricity demand would reach roughly 135 gigawatts. By June, that number had been recalculated at 162 gigawatts — a gain of 27 gigawatts in just four months. To put that in perspective, South Korea operates 26 nuclear reactors totaling about 26 gigawatts. The revision alone is the equivalent of building 19 brand-new large reactors.

The government did not arrive at this figure by revising its GDP growth assumptions. In fact, officials were explicit about that. What changed was the scale and speed of two specific types of industrial investment: semiconductor fabrication and AI datacenters.

This is the first time South Korea’s Electricity Supply and Demand Basic Plan — the legally mandated fifteen-year energy roadmap revised every two years — has been upended so dramatically within a single government cycle. And it is happening just months after a presidential election, which makes the political timing impossible to ignore.

Where the Demand Actually Comes From

The revised forecast attributes most of the 27-gigawatt increase to what planners call “additional demand” — loads from specific industries that sit on top of baseline economic growth projections. The biggest single factor is the semiconductor sector.

Samsung Electronics and SK Hynix together presented a combined demand potential of roughly 40 gigawatts, citing three major projects: the accelerated completion of the Yongin semiconductor cluster, a new four-fab complex planned for the Honam region, and capacity expansions in Pyeongtaek and Cheongju. The planning committee accepted 36.8 gigawatts of that proposal at face value.

Datacenters tell a different story — at least in terms of credibility. Companies had signaled 20.8 gigawatts of planned demand, but the committee reflected only the first phase: 10.8 gigawatts. The second phase, roughly 10 gigawatts, was deferred with the explanation that market conditions and actual project progress need to be watched before locking in commitments.

When adjusted for timing, load profiles, and overlap with baseline growth, the final 2040 additional demand figures came to approximately 24.6 gigawatts for semiconductors and 11.9 gigawatts for datacenters. Together they account for nearly all of the upward revision.

The Selective Trust Problem

Why treat semiconductor demand as near-certain while discounting datacenter demand by nearly half? The committee’s justification rests on the identity of the investors. Samsung and SK Hynix are globally validated champions with construction already underway. “What you can see with your own eyes” was the practical criterion, according to one committee member.

Datacenters, by contrast, operate on a pattern that has haunted previous plan cycles: companies apply for far more grid capacity than they ultimately need. Yoo Seung-hoon, a professor at Seoul Science and Technology University who chaired a subcommittee during earlier plan revisions, pointed out that current nationwide datacenter consumption sits at roughly 1 gigawatt. The question of whether that will multiply fifteen-fold remains unproven.

This asymmetry in treatment creates a genuine accounting opacity. The formula for the final numbers is published. The calibration that turns a 40-gigawatt claim into a 36.8-gigawatt acceptance, or a 20.8-gigawatt datacenter plan into a 10.8-gigawatt inclusion, cannot be independently verified from the materials released so far. Environmental groups have already called for the government to distinguish clearly between confirmed demand and speculative demand before proceeding.

The Supply-Side Answer Is Already Forming

On June 26, the planning committee released its provisional supply outlook. Renewables are expected to grow from the current 37 gigawatts to 220 gigawatts by 2040 — solar dominating at 155 gigawatts, with offshore wind scaling from 0.4 gigawatts to 45 gigawatts. On paper, this is a sixfold expansion.

But generation capacity and actual energy output are not the same thing. Analysts presented at the same session estimated that 100 gigawatts of renewable capacity would produce only about 58 percent of the annual electricity consumed by the mega-projects driving this entire revision. The intermittent nature of solar and wind leaves a structural gap that no amount of gigawatt-targeting closes on its own.

This is exactly why industry is pushing back toward nuclear. Jin Young-hyun, executive vice president of Samsung Electronics, recommended expanded nuclear capacity and LNG combined-heat-and-power plants during a June national report session. The country’s largest electricity consumers are effectively saying the renewable buildout alone cannot power what they are building.

The government’s response has been equally direct. The Climate, Energy and Environment Ministry has formally indicated it is reviewing whether to include new reactors in the next plan and is negotiating to extend the permitted operation period for existing reactors from ten years to twenty years beyond their original design life.

What This Means for South Korea’s Nuclear Industry

South Korea built and exported nuclear reactors for decades — the Barakah plant in the UAE being the most prominent example. That export pipeline stalled under the previous administration’s energy-transition framework, which canceled six planned new reactors and halted lifetime extensions for ten older units. The current revision signals a full reversal of that posture.

Two large reactor sites have already been selected: a pair inyeongbuk province in Yeongdeok and an SMR site in Busan’s Gijang district. Professor Yoo noted that several existing plants — Kori, Youngdeok, Hanbit — have spare land that could shave three to four years off the site-selection process. That time saving is significant in an industry where licensing and construction cycles run a decade or more.

The broader implication extends beyond South Korea’s borders. A domestic nuclear revival creates the industrial and regulatory momentum that sustains reactor exports. Without home demand, the supply chain atrophies. With it, Korea positions itself to re-enter a global market where Japan, France, and China are all competing for orders in regions that need firm, low-carbon baseload power.

The Calculation Nobody Can Escape

The hardest truth about energy planning is that both directions of error are expensive. Overestimate demand and the country builds power plants that sit idle, with costs passed through electricity rates to households and businesses. Underestimate demand and factories cannot get the power they need, which undermines the very semiconductor and datacenter investments driving the revision in the first place.

Minister Kim Sung-hwan acknowledged this directly: accurately forecasting demand through 2040 is “almost impossible.” The pragmatic response, as Professor Yoo framed it, is phased adaptation — build incrementally and respond to actual datacenter growth as it materializes rather than betting the grid on projections.

But there is a precondition attached to that logic. The projected demand must actually materialize. The Honam semiconductor cluster, currently planned with four fabs, is reportedly considering two additional units. If that happens, the 27-gigawatt revision could look conservative rather than aggressive.

The transmission grid presents its own unfunded liability. Currently, KEPCO bears 100 percent of transmission expansion costs, which creates a direct link between demand growth and electricity rate increases. Rail and highway projects receive 30 to 50 percent national budget support. That structural imbalance will determine whether the grid can keep pace with demand or whether rate hikes become the primary constraint on growth.

What Comes Next

The provisional demand figures will be finalized in the government’s October submission, with the full 12th plan expected before year-end. The key unresolved question is whether additional reactors beyond the two large units and one SMR already penciled into the 11th plan will be included — a decision that will be shaped by how convincingly the demand revisions hold up under scrutiny.

The numbers being debated are not abstract. Two hundred megawatts — the power allocated to a single LG Uplus datacenter construction site in Paju — is enough to run tens of thousands of the latest AI chips and equals the annual consumption of 590,000 households. That single facility is the only one of its scale in the capital region. If the 27-gigawatt revision is even partially realized, dozens of facilities like this will need power, and the grid will need to find it somewhere.

South Korea is no longer asking whether it needs nuclear power again. It is asking how fast it can build it.

Who Wins and Who Loses

Semiconductor manufacturers win immediately — they secure the grid capacity their expansion depends on. Existing nuclear operators win through lifetime extensions that improve the economics of their remaining plants. Households and small businesses lose through higher transmission costs and potentially higher electricity rates if demand projections prove accurate and new supply must be built at premium cost.

Renewable developers win on paper — the 220-gigawatt target is ambitious and will generate significant investment flows. But they lose on the timeline, because the intermittent capacity they build cannot replace the firm power that semiconductors and datacenters require without backup, which means nuclear and LNG remain indispensable for the foreseeable future.

Global reactor suppliers watch closely. If South Korea re-establishes a domestic nuclear construction pipeline, it revitalizes the industrial ecosystem — engineering firms, component suppliers, regulatory expertise — that export orders depend on. The window is open. Whether Korea walks through it will be decided by the plan finalized this year.